# MoDiP raw-data profile Generated by `src/profile_data.py` from `data/raw/modip_records.json` (the exact records fetched from the Museum Data Service). - **Records:** 11,865 - **Distinct field types:** 22 - **Distinct material strings:** 476 - **Distinct technique strings:** 82 - **Distinct object-name strings:** 955 - **Distinct use-domain concepts:** 22 ## Field coverage (assertions across the set) | field | assertions | |---|---| | material | 35,172 | | object_name | 28,448 | | dimension | 24,006 | | colour | 16,966 | | object_production_organisation | 15,230 | | organisations_association | 15,230 | | object_name_type | 14,114 | | associated_concept | 13,897 | | brief_description | 12,288 | | license | 11,865 | | license_url | 11,865 | | object_number | 11,865 | | title | 11,865 | | technique | 11,482 | | object_production_place | 9,180 | | inscription_position | 8,271 | | object_production_person | 7,152 | | persons_association | 7,152 | | inscription_content | 7,072 | | inscription_method | 6,840 | | technique_type | 6,047 | | object_production_date | 5,506 | ## The synonymy problem (why a taxonomy is needed) The same polymer is recorded under an abbreviation, a full chemical name and one or more trade names, as independent free-text strings. A keyword search for one misses the others. Witnesses: | polymer | strings that co-refer (with counts) | |---|---| | polypropylene | `PP` (941) + `polypropylene` (940) | | polystyrene | `PS` (843) + `polystyrene` (838) | | polyethylene | `PE` (790) + `polyethylene` (831) + `polythene` (767) | | phenol formaldehyde | `PF` (516) + `phenol formaldehyde` (517) + `bakelite - generic term` (490) | | acrylic (PMMA) | `PMMA` (355) + `polymethyl methacrylate` (352) + `acrylic` (352) + `Perspex - trade name` (14) | | polyamide | `PA` (428) + `polyamide` (431) + `nylon` (426) | Trade-name strings account for **497** material assertions across 151 distinct brand tags (Perspex, Bakelite, Lycra, Neoprene, Teflon, …), each of which resolves to a generic polymer in the taxonomy. ## Free-text cross-references (the latent graph) **323** descriptions contain an inline reference to another object's accession number ("AIBDC 007661.2 is the same box but a different colourway"). These are relationships trapped in prose; `build_graph.py` resolves them into typed object-to-object edges.